"My cake does not look as advertised in the recipe", with this sloppy sentence I hope you understand that you need to give people details, like code, plots, anything. This here is not enough to give any help.

The details about this question:

I have 4 samples with 2 conditions( ctrl_1, ctrl_2, ko_1, ko_2), i use these 4 samples as input. Here is my code:

1 library

```{r setup, include=FALSE} library(DiffBind)

# 2 create object

## 2.1 generate sample_info table
```{r}
# generate columns for sample info.
SampleID <- c("ctrl_1", "ctrl_2", "ko_1", "ko_2")
ControlID <- rep(NA, length(SampleID))
Tissue <- rep("tissue", length(SampleID))
Factor <- rep("accessible_regions", length(SampleID))
Condition <- c("ctrl", "ctrl", "ko", "ko")
Treatment <- c("ctrl", "ctrl", "ko", "ko")
Replicate <- c(rep(1:2, each = 1, time=1))
bamReads <- c("path_to_ctrl_subsample.bam", 
              "path_to_ctrl_subsample.bam", 
              "path_to_ko_subsample.bam",
              "path_to_ko_rmBlack.bam")

Peaks <- c("ctrl_1_peaks.narrowPeak", "ctrl_2_peaks.narrowPeak", "ko_1_peaks.narrowPeak", "ko_2_peaks.narrowPeak")

bamControl <- rep(NA, length(SampleID))
PeakCaller <- rep("bed", length(SampleID))

# combine these columns into one data frame.
sample_sheet <- data.frame(SampleID, ControlID, Tissue, Factor, Condition, Treatment, Replicate, bamReads, Peaks, bamControl,  PeakCaller)

2.2 create DBA object

diff_bind_obj <- dba(sampleSheet=sample_sheet
                     )

diff_bind_obj <-  
  dba.count(diff_bind_obj,
            bUseSummarizeOverlaps = TRUE, # more accurate and slower
            )

diff_bind_obj <- dba.contrast(diff_bind_obj,
             contrast = c("Treatment", "ko", "ctrl"),  # here we have only 2 groups, just assign "ko" vs "ctrl".
             minMembers = 2
             )

diff_bind_obj <- dba.analyze(diff_bind_obj, method=DBA_DESEQ2)

Then, the result is no significannt DARs found. With all Differential Accessible regions' padj =0.9976.

Additionally, i checked my bigwig in igv, it shows that the ctrl_1 is more like ko_1, and ctrl_2's track is more like ko_2's track.

Did anyone have met this situation when doing ATAC-seq to between different conditions, such as control and knock out conditions?